Linus Torvalds [Thu, 18 Aug 2016 22:04:53 +0000 (15:04 -0700)]
Merge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull perf fixes from Ingo Molnar:
"Mostly tooling fixes, but also start/stop filter related fixes, a perf
event read() fix, a fix uncovered by fuzzing, and an uprobes leak fix"
* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
perf/core: Check return value of the perf_event_read() IPI
perf/core: Enable mapping of the stop filters
perf/core: Update filters only on executable mmap
perf/core: Fix file name handling for start/stop filters
perf/core: Fix event_function_local()
uprobes: Fix the memcg accounting
perf intel-pt: Fix occasional decoding errors when tracing system-wide
tools: Sync kvm related header files for arm64 and s390
perf probe: Release resources on error when handling exit paths
perf probe: Check for dup and fdopen failures
perf symbols: Fix annotation of objects with debuginfo files
perf script: Don't disable use_callchain if input is pipe
perf script: Show proper message when failed list scripts
perf jitdump: Add the right header to get the major()/minor() definitions
perf ppc64le: Fix build failure when libelf is not present
perf tools mem: Fix -t store option for record command
perf intel-pt: Fix ip compression
Linus Torvalds [Thu, 18 Aug 2016 18:17:13 +0000 (11:17 -0700)]
Merge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Pull arm64 fixes from Catalin Marinas:
- Avoid a literal load with the MMU off on the CPU resume path
(potential inconsistency between cache and RAM)
- Build error with CONFIG_ACPI=n fixed
- Compiler warning in the arch/arm64/mm/dump.c code fixed
* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
arm64: Fix shift warning in arch/arm64/mm/dump.c
arm64: kernel: avoid literal load of virtual address with MMU off
arm64: Fix NUMA build error when !CONFIG_ACPI
Linus Torvalds [Thu, 18 Aug 2016 18:09:43 +0000 (11:09 -0700)]
Merge tag 'pm-4.8-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Pull power management fixes from Rafael Wysocki:
"More hibernation-related material: one fix for a recent regression in
the core, one small cleanup of the x86-64 resume code and a
documentation update.
Specifics:
- Fix a hibernate core regression resulting from uncovering a latent
bug in its implementation of memory bitmaps by a recent commit
(James Morse).
- Use __pa() to compute a physical address in the x86-64 code
finalizing resume from hibernation (Rafael Wysocki).
- Update power management documentation related to system sleep
states to remove outdated information from it and to add a
description of a recently introduced hibernation debug feature to
it (Rafael Wysocki)"
* tag 'pm-4.8-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
PM / hibernate: Fix rtree_next_node() to avoid walking off list ends
x86/power/64: Use __pa() for physical address computation
PM / sleep: Update some system sleep documentation
Linus Torvalds [Thu, 18 Aug 2016 17:58:50 +0000 (10:58 -0700)]
Merge tag 'drm-fixes-for-4.8-rc3' of git://people.freedesktop.org/~airlied/linux
Pull drm fixes from Dave Airlie:
"Pretty quiet so far:
- a few amdgpu/radeon fixup for pcie pm changes
- a couple of amdgpu fixes
- some build fixes
- printk fix"
* tag 'drm-fixes-for-4.8-rc3' of git://people.freedesktop.org/~airlied/linux:
drm/amdgpu: Change GART offset to 64-bit
drm/mediatek: add ARM_SMCCC dependency
drm/mediatek: add CONFIG_OF dependency
drm/mediatek: add COMMON_CLK dependency
drm/amdgpu: Fix memory trashing if UVD ring test fails
drm/amdgpu: fix vm init error path
drm/amdkfd: print doorbell offset as a hex value
Revert "drm/radeon: work around lack of upstream ACPI support for D3cold"
Revert "drm/amdgpu: work around lack of upstream ACPI support for D3cold"
Johannes Berg [Thu, 11 Aug 2016 09:50:22 +0000 (11:50 +0200)]
locking/barriers: Suppress sparse warnings in lockless_dereference()
After Peter's commit:
331b6d8c7afc ("locking/barriers: Validate lockless_dereference() is used on a pointer type")
... we get a lot of sparse warnings (one for every rcu_dereference, and more)
since the expression here is assigning to the wrong address space.
Instead of validating that 'p' is a pointer this way, instead make
it fail compilation when it's not by using sizeof(*(p)). This will
not cause any sparse warnings (tested, likely since the address
space is irrelevant for sizeof), and will fail compilation when
'p' isn't a pointer type.
Johannes Berg [Thu, 11 Aug 2016 09:50:21 +0000 (11:50 +0200)]
Revert "drm/fb-helper: Reduce READ_ONCE(master) to lockless_dereference"
This reverts commit:
fa7d81bb3c269 ("drm/fb-helper: Reduce READ_ONCE(master) to lockless_dereference")
As Peter explained:
[...] lockless_dereference() is _stronger_ than READ_ONCE(), not weaker.
[...]
Also, clue is in the name: 'dereference', you don't actually dereference
the pointer here, only load it.
My next patch breaks the compile without this revert, because it assumes
you want to deference and thus also need the struct type visible (which
it isn't here), so revert it.
On an i5 laptop, 4 pCPUs, 4vCPUs for one full dynticks guest, there are four
CPU hog processes(for loop) running in the guest, I hot-unplug the pCPUs
on host one by one until there is only one left, then observe CPU utilization
via 'top' in the guest, it shows:
100% st for cpu0(housekeeping)
75% st for other CPUs (nohz full mode)
However, w/o this commit it shows the correct 75% for all four CPUs.
When a guest is interrupted for a longer amount of time, missed clock ticks
are not redelivered later. Because of that, we should not limit the amount
of steal time accounted to the amount of time that the calling functions
think have passed.
However, the interval returned by account_other_time() is NOT rounded down
to the nearest jiffy, while the base interval in get_vtime_delta() it is
subtracted from is, so the max cputime limit is required to avoid underflow.
This patch fixes the regression by limiting the account_other_time() from
get_vtime_delta() to avoid underflow, and lets the other three call sites
(in account_other_time() and steal_account_process_time()) account however
much steal time the host told us elapsed.
perf/core: Check return value of the perf_event_read() IPI
The call to smp_call_function_single in perf_event_read() may fail if
an invalid or not online CPU index is passed. Warn user if such bug is
present and return error.
At this time the perf_addr_filter_needs_mmap() function will _not_
return true on a user space 'stop' filter. But stop filters need
exactly the same kind of mapping that range and start filters get.
Function perf_event_mmap() is called by the MM subsystem each time
part of a binary is loaded in memory. There can be several mapping
for a binary, many times unrelated to the code section.
Each time a section of a binary is mapped address filters are
updated, event when the map doesn't pertain to the code section.
The end result is that filters are configured based on the last map
event that was received rather than the last mapping of the code
segment.
For example if we have an executable 'main' that calls library
'libcstest.so.1.0', and that we want to collect traces on code
that is in that library. The perf cmd line for this scenario
would be:
perf record -e cs_etm// --filter 'filter 0x72c/0x40@/opt/lib/libcstest.so.1.0' --per-thread ./main
Before 'main()' can execute 'libcstest.so.1.0' has to be loaded in
memory. Once that has been done perf_event_mmap() has been called
4 times, with the last map starting at address 0x7fa25ce000 and
the address filter configured to start filtering when the
IP has passed over address 0x0x7fa25ce72c (0x7fa25ce000 + 0x72c).
But that is wrong since the code segment for library 'libcstest.so.1.0'
as been mapped at 0x7fa25bd000, resulting in traces not being
collected.
This patch corrects the situation by requesting that address
filters be updated only if the mapped event is for a code
segment.
perf/core: Fix file name handling for start/stop filters
Binary file names have to be supplied for both range and start/stop
filters but the current code only processes the filename if an
address range filter is specified. This code adds processing of
the filename for start/stop filters.
The reason for the panic is wrong value of __max_logical_packages,
which lets logical_package_map uninitialized and the uncore code
relying on this map being properly initialized (maybe we should
add some safety checks there as well).
The __max_logical_packages is computed as:
DIV_ROUND_UP(total_cpus, ncpus);
- ncpus being number of cores
With above BIOS setup we get total_cpus == 16 which set
__max_logical_packages to 2 (ncpus is 12).
Once topology_update_package_map processes CPU with logical
pkg over 2 we display above messages and fail to initialize
the physical_to_logical_pkg map, which makes the uncore code
crash.
The fix is to remove logical_package_map bitmap completely
and keep and update the logical_packages number instead.
After we enumerate all the present CPUs, we check if the
enumerated logical packages count is within its computed
maximum from BIOS data.
If it's not the case, we set this maximum to the new enumerated
value and freeze any new addition of logical packages.
The freeze is because lot of init code like uncore/rapl/cqm
depends on having maximum logical package value set to allocate
their data, so we can't change it later on.
Prarit Bhargava tested the patch and confirms that it solves
the problem:
Oleg Nesterov [Wed, 17 Aug 2016 15:36:29 +0000 (17:36 +0200)]
uprobes: Fix the memcg accounting
__replace_page() wronlgy calls mem_cgroup_cancel_charge() in "success" path,
it should only do this if page_check_address() fails.
This means that every enable/disable leads to unbalanced mem_cgroup_uncharge()
from put_page(old_page), it is trivial to underflow the page_counter->count
and trigger OOM.
* pm-sleep:
PM / hibernate: Fix rtree_next_node() to avoid walking off list ends
x86/power/64: Use __pa() for physical address computation
PM / sleep: Update some system sleep documentation
1) Buffers powersave frame test is reversed in cfg80211, fix from Felix
Fietkau.
2) Remove bogus WARN_ON in openvswitch, from Jarno Rajahalme.
3) Fix some tg3 ethtool logic bugs, and one that would cause no
interrupts to be generated when rx-coalescing is set to 0. From
Satish Baddipadige and Siva Reddy Kallam.
4) QLCNIC mailbox corruption and napi budget handling fix from Manish
Chopra.
5) Fix fib_trie logic when walking the trie during /proc/net/route
output than can access a stale node pointer. From David Forster.
6) Several sctp_diag fixes from Phil Sutter.
7) PAUSE frame handling fixes in mlxsw driver from Ido Schimmel.
8) Checksum fixup fixes in bpf from Daniel Borkmann.
9) Memork leaks in nfnetlink, from Liping Zhang.
10) Use after free in rxrpc, from David Howells.
11) Use after free in new skb_array code of macvtap driver, from Jason
Wang.
12) Calipso resource leak, from Colin Ian King.
13) mediatek bug fixes (missing stats sync init, etc.) from Sean Wang.
14) Fix bpf non-linear packet write helpers, from Daniel Borkmann.
15) Fix lockdep splats in macsec, from Sabrina Dubroca.
16) hv_netvsc bug fixes from Vitaly Kuznetsov, mostly to do with VF
handling.
17) Various tc-action bug fixes, from CONG Wang.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (116 commits)
net_sched: allow flushing tc police actions
net_sched: unify the init logic for act_police
net_sched: convert tcf_exts from list to pointer array
net_sched: move tc offload macros to pkt_cls.h
net_sched: fix a typo in tc_for_each_action()
net_sched: remove an unnecessary list_del()
net_sched: remove the leftover cleanup_a()
mlxsw: spectrum: Allow packets to be trapped from any PG
mlxsw: spectrum: Unmap 802.1Q FID before destroying it
mlxsw: spectrum: Add missing rollbacks in error path
mlxsw: reg: Fix missing op field fill-up
mlxsw: spectrum: Trap loop-backed packets
mlxsw: spectrum: Add missing packet traps
mlxsw: spectrum: Mark port as active before registering it
mlxsw: spectrum: Create PVID vPort before registering netdevice
mlxsw: spectrum: Remove redundant errors from the code
mlxsw: spectrum: Don't return upon error in removal path
i40e: check for and deal with non-contiguous TCs
ixgbe: Re-enable ability to toggle VLAN filtering
ixgbe: Force VLNCTRL.VFE to be set in all VMDq paths
...
Roman Mashak [Sun, 14 Aug 2016 05:35:02 +0000 (22:35 -0700)]
net_sched: allow flushing tc police actions
The act_police uses its own code to walk the
action hashtable, which leads to that we could
not flush standalone tc police actions, so just
switch to tcf_generic_walker() like other actions.
WANG Cong [Sun, 14 Aug 2016 05:35:01 +0000 (22:35 -0700)]
net_sched: unify the init logic for act_police
Jamal reported a crash when we create a police action
with a specific index, this is because the init logic
is not correct, we should always create one for this
case. Just unify the logic with other tc actions.
WANG Cong [Sun, 14 Aug 2016 05:35:00 +0000 (22:35 -0700)]
net_sched: convert tcf_exts from list to pointer array
As pointed out by Jamal, an action could be shared by
multiple filters, so we can't use list to chain them
any more after we get rid of the original tc_action.
Instead, we could just save pointers to these actions
in tcf_exts, since they are refcount'ed, so convert
the list to an array of pointers.
The "ugly" part is the action API still accepts list
as a parameter, I just introduce a helper function to
convert the array of pointers to a list, instead of
relying on the C99 feature to iterate the array.
WANG Cong [Sun, 14 Aug 2016 05:34:57 +0000 (22:34 -0700)]
net_sched: remove an unnecessary list_del()
This list_del() for tc action is not needed actually,
because we only use this list to chain bulk operations,
therefore should not be carried for latter operations.
David S. Miller [Wed, 17 Aug 2016 23:20:24 +0000 (19:20 -0400)]
Merge branch '1GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net-queue
Jeff Kirsher says:
====================
Intel Wired LAN Driver Updates 2016-08-16
This series contains fixes to e1000e, igb, ixgbe and i40e.
Kshitiz Gupta provides a fix for igb to resolve the PHY delay compensation
math in several functions.
Jarod Wilson provides a fix for e1000e which had to broken up into 2
patches, first is prepares the driver for expanding the list of NICs
that have occasional ~10 hour clock jumps when being used for PTP.
Second patch actually fixes i218 silicon which has been experiencing
the clock jumps while using PTP.
Alex provides 2 patches for ixgbe now that he is back at Intel. First
fixes setting VLNCTRL.VFE bit, which was left unchanged in earlier patches
which resulted in disabling VLAN filtering for all the VFs. Second
corrects the support for disabling the VLAN tag filtering via the
feature bit.
Lastly, David fixes i40e which was causing a kernel panic when
non-contiguous traffic classes or traffic classes not starting with TC0,
were configured on a link partner switch. To fix this, changed the
logic when determining the total number of TCs enabled.
====================
David S. Miller [Wed, 17 Aug 2016 23:18:34 +0000 (19:18 -0400)]
Merge branch 'mlxsw-fixes'
Jiri Pirko says:
====================
mlxsw: IPv4 UC router fixes
Ido says:
Patches 1-3 fix a long standing problem in the driver's init sequence,
which manifests itself quite often when routing daemons try to configure
an IP address on registered netdevs that don't yet have an associated
vPort.
Patches 4-9 add missing packet traps for the router to work properly and
also fix ordering issue following the recent changes to the driver's init
sequence.
The last patch isn't related to the router, but fixes a general problem
in which under certain conditions packets aren't trapped to CPU.
v1->v2:
- Change order of patch 7
- Add patch 6 following Ilan's comment
- Add patchset name and cover letter
====================
Ido Schimmel [Wed, 17 Aug 2016 14:39:37 +0000 (16:39 +0200)]
mlxsw: spectrum: Allow packets to be trapped from any PG
When packets enter the device they are classified to a priority group
(PG) buffer based on their PCP value. After their egress port and
traffic class are determined they are moved to the switch's shared
buffer and await transmission, if:
Packets scheduled to transmission through CPU port (trapped to CPU) use
traffic class 7, which has a zero maximum and minimum quotas. However,
when such packets arrive from PG 0 they are admitted to the shared
buffer as PG 0 has a non-zero minimum quota.
Allow all packets to be trapped to the CPU - regardless of the PG they
were classified to - by assigning a 10KB minimum quota for CPU port and
TC7.
Ido Schimmel [Wed, 17 Aug 2016 14:39:36 +0000 (16:39 +0200)]
mlxsw: spectrum: Unmap 802.1Q FID before destroying it
Before destroying the 802.1Q FID we should first remove the VID-to-FID
mapping. This makes mlxsw_sp_fid_destroy() symmetric with regards to
mlxsw_sp_fid_create().
Fixes: 14d39461b3f4 ("mlxsw: spectrum: Use per-FID struct for the VLAN-aware bridge") Signed-off-by: Ido Schimmel <[email protected]> Signed-off-by: Jiri Pirko <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Ido Schimmel [Wed, 17 Aug 2016 14:39:35 +0000 (16:39 +0200)]
mlxsw: spectrum: Add missing rollbacks in error path
While going over the code I noticed we are missing two rollbacks in the
port's creation error path. Add them and adjust the place of one of them
in the port's removal sequence so that both are symmetric.
Fixes: 56ade8fe3fe1 ("mlxsw: spectrum: Add initial support for Spectrum ASIC") Signed-off-by: Ido Schimmel <[email protected]> Signed-off-by: Jiri Pirko <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Ido Schimmel [Wed, 17 Aug 2016 14:39:33 +0000 (16:39 +0200)]
mlxsw: spectrum: Trap loop-backed packets
One of the conditions to generate an ICMP Redirect Message is that "the
packet is being forwarded out the same physical interface that it was
received from" (RFC 1812).
Therefore, we need to be able to trap such packets and let the kernel
decide what to do with them.
For each RIF, enable the loop-back filter, which will raise the LBERROR
trap whenever the ingress RIF equals the egress RIF.
Ido Schimmel [Wed, 17 Aug 2016 14:39:31 +0000 (16:39 +0200)]
mlxsw: spectrum: Mark port as active before registering it
Commit bbf2a4757b30 ("mlxsw: spectrum: Initialize ports at the end of
init sequence") moved ports initialization to the end of the init
sequence, which means ports are the first to be removed during fini.
Since the FDB delayed work is still active when ports are removed it's
possible for it to process FDB notifications of inactive ports,
resulting in a warning message.
Fix that by marking ports as inactive only after unregistering them. The
NETDEV_UNREGISTER event will invoke bridge's driver port removal
sequence that will cause the FDB (and FDB notifications) to be flushed.
Fixes: bbf2a4757b30 ("mlxsw: spectrum: Initialize ports at the end of init sequence") Signed-off-by: Ido Schimmel <[email protected]> Signed-off-by: Jiri Pirko <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Ido Schimmel [Wed, 17 Aug 2016 14:39:30 +0000 (16:39 +0200)]
mlxsw: spectrum: Create PVID vPort before registering netdevice
After registering a netdevice it's possible for user space applications
to configure an IP address on it. From the driver's perspective, this
means a router interface (RIF) should be created for the PVID vPort.
Therefore, we must create the PVID vPort before registering the
netdevice.
Fixes: 99724c18fc66 ("mlxsw: spectrum: Introduce support for router interfaces") Signed-off-by: Ido Schimmel <[email protected]> Signed-off-by: Jiri Pirko <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Ido Schimmel [Wed, 17 Aug 2016 14:39:29 +0000 (16:39 +0200)]
mlxsw: spectrum: Remove redundant errors from the code
Currently, when device configuration fails we emit errors to the kernel
log despite the fact we already get these from the EMAD transaction
layer, so remove them.
In addition to being unnecessary, removing these error messages will
allow us to reuse mlxsw_sp_port_add_vid() to create the PVID vPort
before registering the netdevice.
Fixes: 99724c18fc66 ("mlxsw: spectrum: Introduce support for router interfaces") Signed-off-by: Ido Schimmel <[email protected]> Signed-off-by: Jiri Pirko <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Ido Schimmel [Wed, 17 Aug 2016 14:39:28 +0000 (16:39 +0200)]
mlxsw: spectrum: Don't return upon error in removal path
When removing a VLAN filter from the device we shouldn't return upon the
first error we encounter, as otherwise we'll have resources that will
never be freed nor used.
Instead, we should keep trying to free as much resources as possible in
a best effort mode.
Remove the error message as well, since we already get these from the
EMAD transaction code.
Fixes: 99724c18fc66 ("mlxsw: spectrum: Introduce support for router interfaces") Signed-off-by: Ido Schimmel <[email protected]> Signed-off-by: Jiri Pirko <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Linus Torvalds [Wed, 17 Aug 2016 19:10:22 +0000 (12:10 -0700)]
Merge tag 'for-v4.8-rc' of git://git.kernel.org/pub/scm/linux/kernel/git/sre/linux-power-supply
Pull power supply fixes from Sebastian Reichel.
* tag 'for-v4.8-rc' of git://git.kernel.org/pub/scm/linux/kernel/git/sre/linux-power-supply:
power_supply: tps65217-charger: fix missing platform_set_drvdata()
power: reset: hisi-reboot: Unmap region obtained by of_iomap
power: reset: reboot-mode: fix build error of missing ioremap/iounmap on UM
power: supply: max17042_battery: fix model download bug.
Ard Biesheuvel [Wed, 17 Aug 2016 15:54:41 +0000 (17:54 +0200)]
arm64: kernel: avoid literal load of virtual address with MMU off
Literal loads of virtual addresses are subject to runtime relocation when
CONFIG_RELOCATABLE=y, and given that the relocation routines run with the
MMU and caches enabled, literal loads of relocated values performed with
the MMU off are not guaranteed to return the latest value unless the
memory covering the literal is cleaned to the PoC explicitly.
So defer the literal load until after the MMU has been enabled, just like
we do for primary_switch() and secondary_switch() in head.S.
Catalin Marinas [Mon, 15 Aug 2016 15:33:10 +0000 (16:33 +0100)]
arm64: Fix NUMA build error when !CONFIG_ACPI
Since asm/acpi.h is only included by linux/acpi.h when CONFIG_ACPI is
enabled, disabling the latter leads to the following build error on
arm64:
arch/arm64/mm/numa.c: In function ‘arm64_numa_init’:
arch/arm64/mm/numa.c:395:24: error: ‘arm64_acpi_numa_init’ undeclared (first use in this function)
if (!acpi_disabled && !numa_init(arm64_acpi_numa_init))
This patch include the asm/acpi.h explicitly in arch/arm64/mm/numa.c for
the arm64_acpi_numa_init() definition.
Fixes: d8b47fca8c23 ("arm64, ACPI, NUMA: NUMA support based on SRAT and SLIT") Reviewed-by: Hanjun Guo <[email protected]> Signed-off-by: Catalin Marinas <[email protected]>
dm raid: support raid0 with missing metadata devices
The raid0 MD personality does not start a raid0 array with any of its
data devices missing.
dm-raid was removing data/metadata device pairs unconditionally if it
failed to read a superblock off the respective metadata device of such
pair, resulting in failure to start arrays with the raid0 personality.
Avoid removing any data/metadata device pairs in case of raid0
(e.g. lvm2 segment type 'raid0_meta') thus allowing MD to start the
array.
Paul Gortmaker [Wed, 17 Aug 2016 10:21:35 +0000 (12:21 +0200)]
clocksource/drivers/mips-gic-timer: Make gic_clocksource_of_init() return int
In commit:
d8152bf85d2c0 ("clocksource/drivers/mips-gic-timer: Convert init function to return error")
several return values were added to a void function resulting in the following warnings:
clocksource/mips-gic-timer.c: In function 'gic_clocksource_of_init':
clocksource/mips-gic-timer.c:175:3: warning: 'return' with a value, in function returning void [enabled by default]
clocksource/mips-gic-timer.c:183:4: warning: 'return' with a value, in function returning void [enabled by default]
clocksource/mips-gic-timer.c:190:3: warning: 'return' with a value, in function returning void [enabled by default]
clocksource/mips-gic-timer.c:195:3: warning: 'return' with a value, in function returning void [enabled by default]
clocksource/mips-gic-timer.c:200:3: warning: 'return' with a value, in function returning void [enabled by default]
clocksource/mips-gic-timer.c:211:2: warning: 'return' with a value, in function returning void [enabled by default]
clocksource/mips-gic-timer.c: At top level:
clocksource/mips-gic-timer.c:213:1: warning: comparison of distinct pointer types lacks a cast [enabled by default]
clocksource/mips-gic-timer.c: In function 'gic_clocksource_of_init':
clocksource/mips-gic-timer.c:183:18: warning: ignoring return value of 'PTR_ERR', declared with attribute warn_unused_result [-Wunused-result]
Given that the addition of the return values was intentional, it seems
that the conversion of the containing function from void to int was
simply overlooked.
I could not figure out why, but GCC cannot prove that the
kona_timer_init() function always initializes its two outputs,
and we get a warning for the use of the 'lsw' variable later,
which is obviously correct.
drivers/clocksource/bcm_kona_timer.c: In function 'kona_timer_init':
drivers/clocksource/bcm_kona_timer.c:119:13: error: 'lsw' may be used uninitialized in this function [-Werror=maybe-uninitialized]
Slightly reordering the loop makes the warning disappear, after
it becomes more obvious to the compiler that the loop is
always entered on the first iteration.
As pointed out by Ray Jui, there is a related problem in the
way we deal with the loop running into the limit, as we just
keep going there with an invalid counter data, so instead we
now propagate a -ETIMEDOUT result to the caller.
Gregory CLEMENT [Wed, 17 Aug 2016 10:21:33 +0000 (12:21 +0200)]
clocksource/drivers/time-armada-370-xp: Fix the clock reference
While converting the init function to return an error, the wrong clock
was get. This leads to the wrong clock rate and slows down the kernel.
For example, it affects typical boot time:
- without fix: over 1 minute
- with fix: 15 seconds
Darrick J. Wong [Wed, 17 Aug 2016 01:12:57 +0000 (11:12 +1000)]
xfs: remove OWN_AG rmap when allocating a block from the AGFL
When we're really tight on space, xfs_alloc_ag_vextent_small() can
allocate a block from the AGFL and give it to the caller. Since the
caller is never the AGFL-fixing method, we must remove the OWN_AG
reverse mapping because it will clash with whatever rmap the caller
wants to set up. This bug was discovered by running generic/299
repeatedly.
Linus Torvalds [Tue, 16 Aug 2016 22:51:57 +0000 (15:51 -0700)]
Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
Pull virtio/vhost fixes from Michael Tsirkin:
- test fixes
- a vsock fix
* tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost:
tools/virtio: add dma stubs
vhost/test: fix after swiotlb changes
vhost/vsock: drop space available check for TX vq
ringtest: test build fix
Linus Torvalds [Tue, 16 Aug 2016 22:50:22 +0000 (15:50 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Pull s390 fixes from Martin Schwidefsky:
"A couple of bug fixes, minor cleanup and a change to the default
config"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
s390/dasd: fix failing CUIR assignment under LPAR
s390/pageattr: handle numpages parameter correctly
s390/dasd: fix hanging device after clear subchannel
s390/qdio: avoid reschedule of outbound tasklet once killed
s390/qdio: remove checks for ccw device internal state
s390/qdio: fix double return code evaluation
s390/qdio: get rid of spin_lock_irqsave usage
s390/cio: remove subchannel_id from ccw_device_private
s390/qdio: obtain subchannel_id via ccw_device_get_schid()
s390/cio: stop using subchannel_id from ccw_device_private
s390/config: make the vector optimized crc function builtin
s390/lib: fix memcmp and strstr
s390/crc32-vx: Fix checksum calculation for small sizes
s390: clarify compressed image code path
Dave Chinner [Tue, 16 Aug 2016 22:41:10 +0000 (08:41 +1000)]
iomap: fiemap should honor the FIEMAP_FLAG_SYNC flag
The flag is checked as supported, but then we do an unconditional
sync of the file, regardless of whether the flag is set or not. Make
the sync conditional on having the FIEMAP_FLAG_SYNC flag set.
iomap: remove superflous pagefault_disable from iomap_write_actor
iov_iter_copy_from_user_atomic disables page faults internally, no need to
do it around the call. This also brings the iomap code in line with
the original filemap version.
iomap: remove superflous mark_page_accessed from iomap_write_actor
This catches up with commit 2457ae ("mm: non-atomically mark page
accessed during page cache allocation where possible"), which
moved the initial access marking into the pagecache allocator.
Darrick J. Wong [Tue, 16 Aug 2016 22:31:49 +0000 (08:31 +1000)]
xfs: store rmapbt block count in the AGF
Track the number of blocks used for the rmapbt in the AGF. When we
get to the AG reservation code we need this counter to quickly
make our reservation during mount.
Dave Chinner [Tue, 16 Aug 2016 22:31:33 +0000 (08:31 +1000)]
xfs: don't invalidate whole file on DAX read/write
When we do DAX IO, we try to invalidate the entire page cache held
on the file. This is incorrect as it will trash the entire mapping
tree that now tracks dirty state in exceptional entries in the radix
tree slots.
What we are trying to do is remove cached pages (e.g from reads
into holes) that sit in the radix tree over the range we are about
to write to. Hence we should just limit the invalidation to the
range we are about to overwrite.
xfs: fix bogus space reservation in xfs_iomap_write_allocate
The space reservations was without an explaination in commit
"Add error reporting calls in error paths that return EFSCORRUPTED"
back in 2003. There is no reason to reserve disk blocks in the
transaction when allocating blocks for delalloc space as we already
reserved the space when creating the delalloc extent.
With this fix we stop running out of the reserved pool in
generic/229, which has happened for long time with small blocksize
file systems, and has increased in severity with the new buffered
write path.
[ dchinner: we still need to pass the block reservation into
xfs_bmapi_write() to ensure we don't deadlock during AG selection.
See commit dbd5c8c ("xfs: pass total block res. as total
xfs_bmapi_write() parameter") for more details on why this is
necessary. ]
Brian Foster [Tue, 16 Aug 2016 22:30:28 +0000 (08:30 +1000)]
xfs: don't assert fail on non-async buffers on ioacct decrement
The buffer I/O accounting mechanism tracks async buffers under I/O. As
an optimization, the buffer I/O count is incremented only once on the
first async I/O for a given hold cycle of a buffer and decremented once
the buffer is released to the LRU (or freed).
xfs_buf_ioacct_dec() has an ASSERT() check for an XBF_ASYNC buffer, but
we have one or two corner cases where a buffer can be submitted for I/O
multiple times via different methods in a single hold cycle. If an async
I/O occurs first, the I/O count is incremented. If a sync I/O occurs
before the hold count drops, XBF_ASYNC is cleared by the time the I/O
count is decremented.
Remove the async assert check from xfs_buf_ioacct_dec() as this is a
perfectly valid scenario. For the purposes of I/O accounting, we really
only care about the buffer async state at I/O submission time.
Dave Ertman [Fri, 12 Aug 2016 16:56:32 +0000 (09:56 -0700)]
i40e: check for and deal with non-contiguous TCs
The i40e driver was causing a kernel panic when
non-contiguous Traffic Classes, or Traffic Classes not
starting with TC0, were configured on a link partner switch.
i40e does not support non-contiguous TCs.
To fix this, the patch changes the logic when determining
the total number of TCs enabled. Before, this would use the
highest TC number enabled and assume that all TCs below it were
also enabled. Now, we create a bitmask of enabled TCs and scan
it to determine not only the number of TCs, but also if the set
of enabled TCs starts at zero and is contiguous. If not, then
DCB is disabled by only returning one TC.
dm raid: enhance attempt_restore_of_faulty_devices() to support more devices
attempt_restore_of_faulty_devices() is limited to 64 when it should support
the new maximum of 253 when identifying any failed devices. It clears any
revivable devices via an MD personality hot remove and add cylce to allow
for their recovery.
Address by using existing functions to retrieve and update all failed
devices' bitfield members in the dm raid superblocks on all RAID devices
and check for any devices to clear in it.
Whilst on it, don't call attempt_restore_of_faulty_devices() for any MD
personality not providing disk hot add/remove methods (i.e. raid0 now),
because such personalities don't support reviving of failed disks.
Alexander Duyck [Fri, 12 Aug 2016 16:53:39 +0000 (09:53 -0700)]
ixgbe: Re-enable ability to toggle VLAN filtering
Back when I submitted the GSO code I messed up and dropped the support for
disabling the VLAN tag filtering via the feature bit. This patch
re-enables the use of the NETIF_F_HW_VLAN_CTAG_FILTER to enable/disable the
VLAN filtering independent of toggling promiscuous mode.
Fixes: b83e30104b ("ixgbe/ixgbevf: Add support for GSO partial") Signed-off-by: Alexander Duyck <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
dm raid: fix restoring of failed devices regression
'lvchange --refresh RaidLV' causes a mapped device suspend/resume cycle
aiming at device restore and resync after transient device failures. This
failed because flag RT_FLAG_RS_RESUMED was always cleared in the suspend path,
thus the device restore wasn't performed in the resume path.
Solve by removing RT_FLAG_RS_RESUMED from the suspend path and resume
unconditionally. Also, remove superfluous comment from raid_resume().
Alexander Duyck [Thu, 11 Aug 2016 21:51:56 +0000 (14:51 -0700)]
ixgbe: Force VLNCTRL.VFE to be set in all VMDq paths
When I was adding the code for enabling VLAN promiscuous mode with SR-IOV
enabled I had inadvertently left the VLNCTRL.VFE bit unchanged as I has
assumed there was code in another path that was setting it when we enabled
SR-IOV. This wasn't the case and as a result we were just disabling VLAN
filtering for all the VFs apparently.
Also the previous patches were always clearing CFIEN which was always set
to 0 by the hardware anyway so I am dropping the redundant bit clearing.
Fixes: 16369564915a ("ixgbe: Add support for VLAN promiscuous with SR-IOV") Signed-off-by: Alexander Duyck <[email protected]> Tested-by: Andrew Bowers <[email protected]> Signed-off-by: Jeff Kirsher <[email protected]>
On LVM2 conversions via lvconvert(8), the target keeps mapped devices in
frozen state when requesting RAID devices be resynchronized. This
applies to e.g. adding legs to a raid1 device or taking over from raid0
to raid4 when the rebuild flag's set on the new raid1 legs or the added
dedicated parity stripe.
Linus Torvalds [Tue, 16 Aug 2016 18:49:02 +0000 (11:49 -0700)]
Merge tag 'pinctrl-v4.8-2' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl
Pull pin control fixes from Linus Walleij:
"Here are a few pin control fixes for the v4.8 series, nothing special
about them:
- Add the missing <linux/io.h> header to the Intel Merrifield driver
to get rid of build mess.
- Drop two instances of pinctrl_unregister() called for drivers using
devm_* resource management.
- Remove the default debounce time for the AMD driver"
* tag 'pinctrl-v4.8-2' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl:
pinctrl: intel: merrifield: Add missed header
pinctrl/amd: Remove the default de-bounce time
pinctrl: pistachio: Drop pinctrl_unregister for devm_ registered device
pinctrl: meson: Drop pinctrl_unregister for devm_ registered device
Ingo Molnar [Tue, 16 Aug 2016 18:08:56 +0000 (20:08 +0200)]
Merge tag 'perf-urgent-for-mingo-20160815' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/urgent
Pull perf/urgent fixes from Arnaldo Carvalho de Melo:
- Fix occasional decoding errors when tracing system-wide with
Intel PT (Adrian Hunter)
- Fix ip compression in Intel PT for some specific packet types not
present on current hardware (Adrian Hunter)
- Fix annotation of objects with debuginfo files (Anton Blanchard)
- Fix build on Fedora Rawhide (25) wrt using the right header to
get the major() & minor() definitions in the jitdump code, now
it is deprecated getting those using sys/types.h, one has to use
sys/sysmacros.h (Arnaldo Carvalho de Melo)
- Sync arm64/s390 kvm related header files (Arnaldo Carvalho de Melo)
- Check for dup and fdopen failures in 'perf probe' (Colin Ian King,
Arnaldo Carvalho de Melo)
- Fix showing callchains in pipe mode, i.e.
perf record -g -o - workload | perf script
now shows callchains (He Kuang)
- Show proper message when the scripts directory points to some
invalid location in 'perf script --list' (He Kuang)
- Fix 'perf mem -t store' to record 'cpu/mem-stores/P' events
again (Jiri Olsa)
- Fix ppc64le build failure when libelf is not present (Ravi Bangoria)
Jarod Wilson [Tue, 26 Jul 2016 18:25:35 +0000 (14:25 -0400)]
e1000e: fix PTP on e1000_pch_lpt variants
I've got reports that the Intel I-218V NIC in Intel NUC5i5RYH systems used
as a PTP slave experiences random ~10 hour clock jumps, which are resolved
if the same workaround for the 82574 and 82583 is employed, so set the
appropriate flag2 in e1000_pch_lpt_info too.
Jarod Wilson [Tue, 26 Jul 2016 18:25:34 +0000 (14:25 -0400)]
e1000e: factor out systim sanitization
This is prepatory work for an expanding list of adapter families that have
occasional ~10 hour clock jumps when being used for PTP. Factor out the
sanitization function and convert to using a feature (bug) flag, per
suggestion from Jesse Brandeburg.
Littering functional code with device-specific checks is much messier than
simply checking a flag, and having device-specific init set flags as needed.
There are probably a number of other cases in the e1000e code that
could/should be converted similarly.
igb: fix adjusting PTP timestamps for Tx/Rx latency
Fix PHY delay compensation math in igb_ptp_tx_hwtstamp() and
igb_ptp_rx_rgtstamp. Add PHY delay compensation in
igb_ptp_rx_pktstamp().
In the IGB driver, there are two functions that retrieve timestamps
received by the PHY - igb_ptp_rx_rgtstamp() and igb_ptp_rx_pktstamp().
The previous commit only changed igb_ptp_rx_rgtstamp(), and the change
was incorrect.
There are two instances in which PHY delay compensations should be
made:
- Before the packet transmission over the PHY, the latency between
when the packet is timestamped and transmission of the packets,
should be an add operation, but it is currently a subtract.
- After the packets are received from the PHY, the latency between
the receiving and timestamping of the packets should be a subtract
operation, but it is currently an add.
James Morse [Tue, 16 Aug 2016 09:46:38 +0000 (10:46 +0100)]
PM / hibernate: Fix rtree_next_node() to avoid walking off list ends
rtree_next_node() walks the linked list of leaf nodes to find the next
block of pages in the struct memory_bitmap. If it walks off the end of
the list of nodes, it walks the list of memory zones to find the next
region of memory. If it walks off the end of the list of zones, it
returns false.
This leaves the struct bm_position's node and zone pointers pointing
at their respective struct list_heads in struct mem_zone_bm_rtree.
memory_bm_find_bit() uses struct bm_position's node and zone pointers
to avoid walking lists and trees if the next bit appears in the same
node/zone. It handles these values being stale.
Swap rtree_next_node()s 'step then test' to 'test-next then step',
this means if we reach the end of memory we return false and leave
the node and zone pointers as they were.
This fixes a panic on resume using AMD Seattle with 64K pages:
[ 6.868732] Freezing user space processes ... (elapsed 0.000 seconds) done.
[ 6.875753] Double checking all user space processes after OOM killer disable... (elapsed 0.000 seconds)
[ 6.896453] PM: Using 3 thread(s) for decompression.
[ 6.896453] PM: Loading and decompressing image data (5339 pages)...
[ 7.318890] PM: Image loading progress: 0%
[ 7.323395] Unable to handle kernel paging request at virtual address 00800040
[ 7.330611] pgd = ffff000008df0000
[ 7.334003] [00800040] *pgd=00000083fffe0003, *pud=00000083fffe0003, *pmd=00000083fffd0003, *pte=0000000000000000
[ 7.344266] Internal error: Oops: 96000005 [#1] PREEMPT SMP
[ 7.349825] Modules linked in:
[ 7.352871] CPU: 2 PID: 1 Comm: swapper/0 Tainted: G W I 4.8.0-rc1 #4737
[ 7.360512] Hardware name: AMD Overdrive/Supercharger/Default string, BIOS ROD1002C 04/08/2016
[ 7.369109] task: ffff8003c0220000 task.stack: ffff8003c0280000
[ 7.375020] PC is at set_bit+0x18/0x30
[ 7.378758] LR is at memory_bm_set_bit+0x24/0x30
[ 7.383362] pc : [<ffff00000835bbc8>] lr : [<ffff0000080faf18>] pstate: 60000045
[ 7.390743] sp : ffff8003c0283b00
[ 7.473551]
[ 7.475031] Process swapper/0 (pid: 1, stack limit = 0xffff8003c0280020)
[ 7.481718] Stack: (0xffff8003c0283b00 to 0xffff8003c0284000)
[ 7.800075] Call trace:
[ 7.887097] [<ffff00000835bbc8>] set_bit+0x18/0x30
[ 7.891876] [<ffff0000080fb038>] duplicate_memory_bitmap.constprop.38+0x54/0x70
[ 7.899172] [<ffff0000080fcc40>] snapshot_write_next+0x22c/0x47c
[ 7.905166] [<ffff0000080fe1b4>] load_image_lzo+0x754/0xa88
[ 7.910725] [<ffff0000080ff0a8>] swsusp_read+0x144/0x230
[ 7.916025] [<ffff0000080fa338>] load_image_and_restore+0x58/0x90
[ 7.922105] [<ffff0000080fa660>] software_resume+0x2f0/0x338
[ 7.927752] [<ffff000008083350>] do_one_initcall+0x38/0x11c
[ 7.933314] [<ffff000008b40cc0>] kernel_init_freeable+0x14c/0x1ec
[ 7.939395] [<ffff0000087ce564>] kernel_init+0x10/0xfc
[ 7.944520] [<ffff000008082e90>] ret_from_fork+0x10/0x40
[ 7.949820] Code: d28000228b400c21f98000319ac32043 (c85f7c22)
[ 7.955909] ---[ end trace 0024a5986e6ff323 ]---
[ 7.960529] Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b
Here struct mem_zone_bm_rtree's start_pfn has been returned instead of
struct rtree_node's addr as the node/zone pointers are corrupt after
we walked off the end of the lists during mark_unsafe_pages().
This behaviour was exposed by commit 6dbecfd345a6 ("PM / hibernate:
Simplify mark_unsafe_pages()"), which caused mark_unsafe_pages() to call
duplicate_memory_bitmap(), which uses memory_bm_find_bit() after walking
off the end of the memory bitmap.
Fixes: 3a20cb177961 (PM / Hibernate: Implement position keeping in radix tree) Signed-off-by: James Morse <[email protected]>
[ rjw: Subject ] Signed-off-by: Rafael J. Wysocki <[email protected]>
Jim Lin [Tue, 16 Aug 2016 07:18:05 +0000 (10:18 +0300)]
usb: xhci: Fix panic if disconnect
After a device is disconnected, xhci_stop_device() will be invoked
in xhci_bus_suspend().
Also the "disconnect" IRQ will have ISR to invoke
xhci_free_virt_device() in this sequence.
xhci_irq -> xhci_handle_event -> handle_cmd_completion ->
xhci_handle_cmd_disable_slot -> xhci_free_virt_device
If xhci->devs[slot_id] has been assigned to NULL in
xhci_free_virt_device(), then virt_dev->eps[i].ring in
xhci_stop_device() may point to an invlid address to cause kernel
panic.
virt_dev = xhci->devs[slot_id];
:
if (virt_dev->eps[i].ring && virt_dev->eps[i].ring->dequeue)
Alban Browaeys [Tue, 16 Aug 2016 07:18:04 +0000 (10:18 +0300)]
xhci: really enqueue zero length TRBs.
Enqueue the first TRB even if full_len is zero.
Without this "adb install <apk>" freezes the system.
Signed-off-by: Alban Browaeys <[email protected]> Fixes: 86065c2719a5 ("xhci: don't rely on precalculated value of needed trbs in the enqueue loop") Signed-off-by: Mathias Nyman <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
Mathias Nyman [Tue, 16 Aug 2016 07:18:03 +0000 (10:18 +0300)]
xhci: always handle "Command Ring Stopped" events
Fix "Command completion event does not match command" errors by always
handling the command ring stopped events.
The command ring stopped event is generated as a result of aborting
or stopping the command ring with a register write. It is not caused
by a command in the command queue, and thus won't have a matching command
in the comman list.
Solve it by handling the command ring stopped event before checking for a
matching command.
In most command time out cases we abort the command ring, and get
a command ring stopped event. The events command pointer will point at
the current command ring dequeue, which in most cases matches the timed
out command in the command list, and no error messages are seen.
If we instead get a command aborted event before the command ring stopped
event, the abort event will increse the command ring dequeue pointer, and
the following command ring stopped events command pointer will point at the
next, not yet queued command. This case triggered the error message
net: ethernet: mediatek: fix runtime warning raised by inconsistent struct device pointers passed to DMA API
Runtime warning occurs if DMA-API debug feature is enabled that would be
raised by pointers passed to DMA API as arguments to inconsistent struct
device objects, so that the patch makes them usage aligned between DMA
operations such as dma_map_*() and dma_unmap_*() to eliminate the warning.
net: ethernet: mediatek: fix flow control settings on GMAC0 is not being enabled properly
Commit 08ef55c6f257acf3bdc6940813f80e8f0f5d90ec
("net-next: mediatek: fix gigabit and flow control advertisement")
had supported proper flow control settings for GMAC1. But for GMAC0,
1.GMAC0 shares the common logic with GMAC1 inside mtk_phy_link_adjust()
to adapt various settings for the target phy.
2.GMAC0 uses fixed-phy to connect to a builtin gigabit switch with
fixed link speed as commit 0c72c50f6f93b0c3daa9ea35d89ab3a933c7b5a0
("net-next: mediatek: add fixed-phy support") describes.
3.However, fixed-phy doesn't enable SUPPORTED_Pause & SUPPORTED_Asym_Pause
supported flag on default that would cause mtk_phy_link_adjust() not to
enable flow control setting on GMAC0 properly and cause packet dropped
when high traffic.
Due to these reasons, the patch adds SUPPORTED_Pause & SUPPORTED_Asym_Pause
supported flags on fixed-phy used by the driver to have proper handling on
the both GMAC with the shared common logic.
net: ethernet: mediatek: fix RMII mode and add REVMII supported by GMAC
The patch fixes up the incorrect setup of reduced MII (RMII) on GMAC
and adds the supplement for the setup of reverse MII (REVMII) on GMAC
, and rearranges the error handling for invalid PHY argument.
Adrian Hunter [Mon, 15 Aug 2016 07:23:04 +0000 (10:23 +0300)]
perf intel-pt: Fix occasional decoding errors when tracing system-wide
In order to successfully decode Intel PT traces, context switch events
are needed from the moment the trace starts. Currently that is ensured
by using the 'immediate' flag which enables the switch event when it is
opened.
However, since commit 86c2786994bd ("perf intel-pt: Add support for
PERF_RECORD_SWITCH") that might not always happen. When tracing
system-wide the context switch event is added to the tracking event
which was not set as 'immediate'. Change that so it is.
I feel like we should maybe return -ENOMEM or -ENOBUFS, but I'm not sure
userspace is equipped to handle that. Anyway, this is better than a GPF
and looks somewhat consistent with other tipc_msg_create() callers.
David S. Miller [Mon, 15 Aug 2016 20:48:08 +0000 (13:48 -0700)]
Merge branch 'hv_netvsc-VF-removal-fixes'
Vitaly Kuznetsov says:
====================
hv_netvsc: fixes for VF removal path
Kernel crash is reported after VF is removed and detached from netvsc
device. Turns out we have multiple different (but related) issues on the
VF removal path which I'm trying to address with PATCHes 2-5 of this
series. PATCH1 is required to support the change.
Changes since v1:
- Re-arrange patches in the series to not introduce new issues [David Miller]
- Add PATCH5 which fixes a new issue I discovered while testing.
- Add Haiyang' A-b tags to PATCH1-4
With regards to Stephen's suggestion: I believe that switching to using RCU
and eliminating vf_use_cnt/vf_inject is the right thing to do long-term, we
can either put this on top of this series or do it later in net-next.
====================
Vitaly Kuznetsov [Mon, 15 Aug 2016 15:48:43 +0000 (17:48 +0200)]
hv_netvsc: fix bonding devices check in netvsc_netdev_event()
Bonding driver sets IFF_BONDING on both master (the bonding device) and
slave (the real NIC) devices and in netvsc_netdev_event() we want to skip
master devices only. Currently, there is an uncertainty when a slave
interface is removed: if bonding module comes first in netdev_chain it
clears IFF_BONDING flag on the netdev and netvsc_netdev_event() correctly
handles NETDEV_UNREGISTER event, but in case netvsc comes first on the
chain it sees the device with IFF_BONDING still attached and skips it. As
we still hold vf_netdev pointer to the device we crash on the next inject.
Vitaly Kuznetsov [Mon, 15 Aug 2016 15:48:42 +0000 (17:48 +0200)]
hv_netvsc: protect module refcount by checking net_device_ctx->vf_netdev
We're not guaranteed to see NETDEV_REGISTER/NETDEV_UNREGISTER notifications
only once per VF but we increase/decrease module refcount unconditionally.
Check vf_netdev to make sure we don't take/release it twice. We presume
that only one VF per netvsc device may exist.
Vitaly Kuznetsov [Mon, 15 Aug 2016 15:48:41 +0000 (17:48 +0200)]
hv_netvsc: reset vf_inject on VF removal
We reset vf_inject on VF going down (netvsc_vf_down()) but we don't on
VF removal (netvsc_unregister_vf()) so vf_inject stays 'true' while
vf_netdev is already NULL and we're trying to inject packets into NULL
net device in netvsc_recv_callback() causing kernel to crash.
Vitaly Kuznetsov [Mon, 15 Aug 2016 15:48:40 +0000 (17:48 +0200)]
hv_netvsc: avoid deadlocks between rtnl lock and vf_use_cnt wait
Here is a deadlock scenario:
- netvsc_vf_up() schedules netvsc_notify_peers() work and quits.
- netvsc_vf_down() runs before netvsc_notify_peers() gets executed. As it
is being executed from netdev notifier chain we hold rtnl lock when we
get here.
- we enter while (atomic_read(&net_device_ctx->vf_use_cnt) != 0) loop and
wait till netvsc_notify_peers() drops vf_use_cnt.
- netvsc_notify_peers() starts on some other CPU but netdev_notify_peers()
will hang on rtnl_lock().
- deadlock!
Instead of introducing additional synchronization I suggest we drop
gwrk.dwrk completely and call NETDEV_NOTIFY_PEERS directly. As we're
acting under rtnl lock this is legitimate.
Vitaly Kuznetsov [Mon, 15 Aug 2016 15:48:39 +0000 (17:48 +0200)]
hv_netvsc: don't lose VF information
struct netvsc_device is not suitable for storing VF information as this
structure is being destroyed on MTU change / set channel operation (see
rndis_filter_device_remove()). Move all VF related stuff to struct
net_device_context which is persistent.
Simon Horman [Mon, 15 Aug 2016 11:06:24 +0000 (13:06 +0200)]
gre: set inner_protocol on xmit
Ensure that the inner_protocol is set on transmit so that GSO segmentation,
which relies on that field, works correctly.
This is achieved by setting the inner_protocol in gre_build_header rather
than each caller of that function. It ensures that the inner_protocol is
set when gre_fb_xmit() is used to transmit GRE which was not previously the
case.
I have observed this is not the case when OvS transmits GRE using
lwtunnel metadata (which it always does).
projects / linux.git / log